The present invention relates to storage of data in a multi-tier storage system, and more specifically, to selecting a migration decision window for stored data based upon characteristics of hotspots in the multi-tier enterprise storage system.
In general, multi-tier enterprise storage systems include at least two types of data storage devices that have different performance characteristics. These multi-tier enterprise storage systems are designed to optimize system performance and maximize utilization of the storage devices by migrating hot data to faster storage devices, such as solid state devices (SSDs) and cold data to appropriate storage device types, such as hard disk drives (HDDs). As used herein, the term hot data refers to data that experiences a relatively large number of random accesses and the term cold data refers to data that experiences a relatively large number of sequential accesses or data that experiences a relatively low number of total accesses. In general, hotspots, which are storage areas that contain hot data, in multi-tier enterprise storage systems are dynamic, that is the locations of hotspots change over time. For example, a particular portion of storage may be considered a hotspot for a specific period of time and will not be considered a hotspot during another period of time.
Typically, access statistics for data stored in the multi-tier enterprise storage systems are collected over time and periodically reviewed. These statistics are evaluated at regular intervals, often by an automated algorithm, referred to as a decision window, in order to identify which data is hot and which data is cold. In currently available systems, the decision window is a static system parameter that is set by an administrator. In general, a static decision window may result in sub-optimal system performance and sub-optimal utilization of the storage devices in the multi-tier enterprise storage systems. For example, a decision window which is much longer than the duration of a hotspot may result in poor responsiveness in the system, since the hot data is not placed on a faster storage device in time for the large number of accesses.